Statistical Arb/Pairs trading strategy!

[marijan993]

interesting idea, Mari.

let’s see, firstly, divide EU with GU price:

i’m wondering if this case is mathmatically identical to trade a SMA on EURGBP itself. when eg is much higher than its SMA, we short, and vice versa. i’m not sure…

secondly, substract GU and EU…, now this seems really a gap indicator now. but if this cap is widening(narrowing), how do we know it is due to eur’s crazy move or gbp’s? i guess we still have to use “overlay image”… but this “gap indicator” does provide a quantity tool instead of pure visual decision…

Yes, of course we would still have to use overlay for this.

[marijan993]

You are thinking about having a common SMA for both? with the same value on both charts? That could be good comparation method, but it will not be visible on 1m because the idea is to have a median between EU and GU, that is (EU+GU)/2 and that will give something about 1.4.

With this, EU will be always below that SMA and GU always above. Then you can get the standard deviation using that SMA for the mean and that could leave with something to compare which pair is above that standard deviation and which one is not.

If you can give me an example on Excel on how to use this, I can code it as an indicator for all of us

I’ll try to post an image to explain what I mean.

[marijan993]

This is done with OffChartPrice indicator in sub-window. Yellow line is EURUSD, red line is GBPUSD, and blue line is 50 SMA of EURUSD. Now, if the blue line could be SMA of both EURUSD/GBPUSD, then we could sell EURUSD when itis above SMA, and buy GBPUSD when it is below SMA. Hope I’m making myself clear.

[pipcompounder]

[QUOTE=medisoft;352202]I did some very nice trades on the London’s time. About 6 successful new trades and closed some of the old trades, but not all of them.

I think that maybe trading the pairs which market is open is a good idea, for example EU/GU works pretty good on London’s time, but for me on my daytime it doesn’t work very good, it takes too long to close a trade and sometimes it takes days or weeks.

So for EU/GU maybe the 0:00-4:00 CST is good time because that period of time is when more volatility is (just before and after the market opening) for EUR and GBP, but low volatility on USD…

QUOTE]

I Agree, medisoft, it seems my EG trades start hanging around after 11am CST, to me 1-10am CST seems good for EU\GU pairs…

[medisoft]

Set the blue line to be the SMA of EURGBP

@marijan993

This is done with OffChartPrice indicator in sub-window. Yellow line is EURUSD, red line is GBPUSD, and blue line is 50 SMA of EURUSD. Now, if the blue line could be SMA of both EURUSD/GBPUSD, then we could sell EURUSD when itis above SMA, and buy GBPUSD when it is below SMA. Hope I’m making myself clear.

[marijan993]

Set the blue line to be the SMA of EURGBP

I’ve tried it before, it got me nowhere.

[christinaa]

I’m using the visual kelton’s method to enter the trades on both pairs, and my TP is only 5 pips. That is giving me nice results of more than 1 % daily.

I’m risking 5 % of my account on each trade and that 5 % is reached when the trade has 100 pips loss.

can i ask what time frame are you using when your doing this TP 5 pips, SL 100 pips?

[christinaa]

look to see which pair is below the other…so right now the GU is under the EU, 20 pip gap, so since EG has the Euro as the first, or primary currency, I sell EG…yesterday the EU was above the GU, so I was buying EG…if you put BB on, it gives you a clue, too.

Pipcompounder GU under EU,
is same as EU above GU

can you please clarify for me when to sell or buy the EG

[pipcompounder]

Pipcompounder GU under EU,
is same as EU above GU

can you please clarify for me when to sell or buy the EG

Your right, I meant on the second part GU above EU, or the opposite in other words…sorry!

If the “e” is higher, sell EG, if “G” is higher, buy EG…does that help?

Also, how are your trades doing?

[christinaa]

I have a theory that is only half-way thought through and maybe someone can help complete it by bringing your thoughts to the table…

first, a premise: I trade the 1 min TF, like Kelton, so that I can make short term profits because the gap will close again within a short time period, usually less than 24 hours (yes, yesterday was weird that way). sometimes with hugh moves, the charts realign themselves, which is fine if no trades are open, because then if gap opens it’s accurate for that short time span (less than 24 hours).

ok, now, since EG shows accurate ratio between EU and GU price mathematically, what if the charts could be moved to show the price scales on right to show that ratio…i.e., when EU chart has the price of 1.2700 overlapping the GU price of 1.5800, will this produce accurate “gaps”? this doesn’t mean actual price, just aligning the charts up or down until the GU price 1.5800 showing thru the EU chart, using VItrite is overlaid on top of 1.2700…this may not have benefit at all to this type of trading, but I just thought it might actually have total benefit…when I did this myself, the pairs showed mostly correlated, overlapping each other, with a few gaps…the application of this idea would involve fixing charts and being able to have much head room above and below the price range.

If your worried about the 1M chart realigning itself after a huge move, then place a vertical line on the EURUSD chart and a vertical line on the GBPUSD chart directly on top of the line on the EURUSD chart, when the charts realign and u see 2 vertical lines use the fix to scale to adjust the charts so the vertical lines can be back on top of each other
hope it helps

[marijan993]

[QUOTE=medisoft;352202]I did some very nice trades on the London’s time. About 6 successful new trades and closed some of the old trades, but not all of them.

I think that maybe trading the pairs which market is open is a good idea, for example EU/GU works pretty good on London’s time, but for me on my daytime it doesn’t work very good, it takes too long to close a trade and sometimes it takes days or weeks.

So for EU/GU maybe the 0:00-4:00 CST is good time because that period of time is when more volatility is (just before and after the market opening) for EUR and GBP, but low volatility on USD…

QUOTE]

I Agree, medisoft, it seems my EG trades start hanging around after 11am CST, to me 1-10am CST seems good for EU\GU pairs…

Can you please explain how to post charts when used with virite overlay?

[pipcompounder]

Christiana, Wednesday was one of those days, where a gap appeared, but didn’t really come back…the EG was down below 0.8000, and when it came up to 0.8050, there was about a 60 pip gap which showed the EU above the GU, do good time to sell, but instead that gap increased until EG was up to 0.8035, and hasn’t come below that since, now it closed for the week above 0.8070, so while many are hoping it comes back down all the way to 0.8000, the 1min chart closed showing only 60pip gap again, so if that closes, EG will only come down to about 0.8035…this is making me wonder if using EG is that great of idea, but when I go back and apply medisoft’s idea of only trading gaps for limited time of day, that would have worked fine for this weird week…also, closing trades early instead do waiting for gap to completely close seems to help reduce risk of ever widening gap…now, these gaps most likely will eventually close, i.e., the EG will probably come back down to 0.8000, but when? Maybe in a month? I think how the banks and ECB, and the Fed are playing this euro crisis is affecting the natural mean reversion that makes this strategy work, because correlation itself isn’t enough, we need mean reversion.

[fxbreakz]

Medisoft, re post #664:
This is what I was looking for, not biggest TP (actually I asked about biggest ratio). Are these from a random sample picked since 07 1999? If [U]let’s say[/U] 1188 winners and 334 losers:
%W=78.05, E=0.56, SD=1.24, SQN(100 trades)=4.51, PF=1.77
with 3.7% risked per 10 pips, starting with 1K -> 20B, with 1% / pip -> gazillions
How many day would it take on avg for all these trades?
Thx!

[medisoft]

I’m doing the stats, I hope to have finished on Monday, and the stats aim to have the right TP and SL to close all the trades daily, to not have trades opened for the next day, something to make statarb a 8 hours work hehehehe, but a work that has no boss, no office to go, no suit to dress, and no risk of losing the job because problems in any company or the world And maybe, just maybe, to make a gold mine to share!

Medisoft, re post #664:
This is what I was looking for, not biggest TP (actually I asked about biggest ratio). Are these from a random sample picked since 07 1999? If [U]let’s say[/U] 1188 winners and 334 losers:
%W=78.05, E=0.56, SD=1.24, SQN(100 trades)=4.51, PF=1.77
with 3.7% risked per 10 pips, starting with 1K -> 20B, with 1% / pip -> gazillions
How many day would it take on avg for all these trades?
Thx!

[FXEZ]

Christiana, Wednesday was one of those days, where a gap appeared, but didn’t really come back…the EG was down below 0.8000, and when it came up to 0.8050, there was about a 60 pip gap which showed the EU above the GU, do good time to sell, but instead that gap increased until EG was up to 0.8035, and hasn’t come below that since, now it closed for the week above 0.8070, so while many are hoping it comes back down all the way to 0.8000, the 1min chart closed showing only 60pip gap again, so if that closes, EG will only come down to about 0.8035…this is making me wonder if using EG is that great of idea, but when I go back and apply medisoft’s idea of only trading gaps for limited time of day, that would have worked fine for this weird week…also, closing trades early instead do waiting for gap to completely close seems to help reduce risk of ever widening gap…now, these gaps most likely will eventually close, i.e., the EG will probably come back down to 0.8000, but when? Maybe in a month? I think how the banks and ECB, and the Fed are playing this euro crisis is affecting the natural mean reversion that makes this strategy work, because correlation itself isn’t enough, we need mean reversion.

pipcompounder,

I don’t believe that trading the EG is the cause of the non-reversion, because non-reversion also happens on the individual pairs EU and GU. The thing that I see is that the markets (in this case EUR/USD and GBP/USD) don’t care about retail traders who want to come in, trade for a few hours and grab a bucket of pips, then do it again the next day. The normal divergences of correlated pairs is greater generally than what many expect. Our job as objective observers of the markets is to adjust ourselves and our expectations to the fundamental realities of the actual longer term fluctuations between two correlated symbols like EU and GU. We do this by altering our position sizing to account for intraday 200-300 pip divergences which are “ordinary”, with an eye on the daily charts where 1000 pip divergences are “ordinary”.

Let me put it this way. When we apply 1 minute charts to exploit minor divergences, the strategy often works well for a time. Then over time the longer term divergences come into play because of real weakness / strength in the economies of the underlying currencies and the reaction of the longer term players. Some of these players (funds) persistently buy or sell one against the other and this throws our 1 minute divergence into disarray when we take initial positions that are too large.

The reality of a strategy where adding takes place is that it is a martingale. All the caveats and dangers (risk of ruin) associated with martingale strategies apply, even though the strategy is based on the spread, and not the underlying.

I will say this, however. There is more opportunity for short term mean reversion on the EU and GU than on EG. However, when EG is trending, the spread (EU-GU) will also be trending.

[medisoft]

1 minute, exact Kelton’s technique, without modification, only I defined small take profit and calculate the lot size based on a 100 pip stop loss.

can i ask what time frame are you using when your doing this TP 5 pips, SL 100 pips?

[fxbreakz]

The strategies closest to Holy Grails have a 1/1 Risk:Reward and higher winning prob. (70 or higher) or 1/2 and around 50% prob. of winning. The reason they are better than high risk / low reward with very high winning prob. is because they have low standard deviation, thus higher SQN, which makes huge difference in profit.

[Jedster]

pipcompounder,

I don’t believe that trading the EG is the cause of the non-reversion, because non-reversion also happens on the individual pairs EU and GU. The thing that I see is that the markets (in this case EUR/USD and GBP/USD) don’t care about retail traders who want to come in, trade for a few hours and grab a bucket of pips, then do it again the next day. The normal divergences of correlated pairs is greater generally than what many expect. Our job as objective observers of the markets is to adjust ourselves and our expectations to the fundamental realities of the actual longer term fluctuations between two correlated symbols like EU and GU. We do this by altering our position sizing to account for intraday 200-300 pip divergences which are “ordinary”, with an eye on the daily charts where 1000 pip divergences are “ordinary”.

Let me put it this way. When we apply 1 minute charts to exploit minor divergences, the strategy often works well for a time. Then over time the longer term divergences come into play because of real weakness / strength in the economies of the underlying currencies and the reaction of the longer term players. Some of these players (funds) persistently buy or sell one against the other and this throws our 1 minute divergence into disarray when we take initial positions that are too large.

The reality of a strategy where adding takes place is that it is a martingale. All the caveats and dangers (risk of ruin) associated with martingale strategies apply, even though the strategy is based on the spread, and not the underlying.

I will say this, however. There is more opportunity for short term mean reversion on the EU and GU than on EG. However, when EG is trending, the spread (EU-GU) will also be trending.

Hi all, how’s things going. Enjoying the opportunities?

That is a good post. It goes a long way to explain why I moved from the 1M to the 5M and then onto the 60M. I don;t like the potential risk/exposure trading the smaller timeframes. I have just seen out a very long (relatively speaking) period for a long trade which I took from my 60M chart on the 7th May at 0.8060. It has taken 2 weeks to come back to just break even. At one point it was over 110 pips against me. I am seriously considering moving to the daily charts. Exactly the same principle, but trade the wider correlation movements (that FXEZ is referring to) instead of the intraday ones.

However, as I read my 1H charts at the moment, now does look like a very good opportunity to short the EG. Price passed +2SD, went through to +2.5SD and is back dorwn at the +2SD level. The big down trend is potentially still on, MA (from bollinger bands) is pretty much flat. So, I have a sell stop at 0.8064 (just below lows of the last few hours). If price continues up to +3SD around the 0.81 level, I’ll enter short there instead.

[pipcompounder]

I see what you’re saying, when I pull a 5 min chart compared to 1 min, it shows there was an up to 100 pip gap yesterday, which the EG got up to 0.8090, but now has come down significantly, and before that, earlier in the week it had the GU well above the EU, which was when the EG dipped below 0.8000…this would have been great buy opportunity, but also having trade go for 3-4 days…both of these moves faked me out on 1min chart, honestly, even though I made some good shorter trades during these times…I am re-evaluating this now.

[medisoft]

I returned to the 1 minute timeframe and had various successful trades. What I see while viewing the charts is that sometimes one pair fail out of correlation only 1 minute, moving very fast and on the next minute the prices returned to correlation. This type of fast moves are the ones that 1 minute view exploits very good.

Small take profits and my brand new stop loss make a high probability trades here. And also, doing this only when the currency is open makes it better trades!